KVM: x86: allow TSC to differ by NTP correction bounds without TSC scaling
authorMarcelo Tosatti <mtosatti@redhat.com>
Tue, 16 Jun 2020 11:47:41 +0000 (08:47 -0300)
committerPaolo Bonzini <pbonzini@redhat.com>
Tue, 23 Jun 2020 09:55:17 +0000 (05:55 -0400)
The Linux TSC calibration procedure is subject to small variations
(its common to see +-1 kHz difference between reboots on a given CPU, for example).

So migrating a guest between two hosts with identical processor can fail, in case
of a small variation in calibrated TSC between them.

Without TSC scaling, the current kernel interface will either return an error
(if user_tsc_khz <= tsc_khz) or enable TSC catchup mode.

This change enables the following TSC tolerance check to
accept KVM_SET_TSC_KHZ within tsc_tolerance_ppm (which is 250ppm by default).

        /*
         * Compute the variation in TSC rate which is acceptable
         * within the range of tolerance and decide if the
         * rate being applied is within that bounds of the hardware
         * rate.  If so, no scaling or compensation need be done.
         */
        thresh_lo = adjust_tsc_khz(tsc_khz, -tsc_tolerance_ppm);
        thresh_hi = adjust_tsc_khz(tsc_khz, tsc_tolerance_ppm);
        if (user_tsc_khz < thresh_lo || user_tsc_khz > thresh_hi) {
                pr_debug("kvm: requested TSC rate %u falls outside tolerance [%u,%u]\n", user_tsc_khz, thresh_lo, thresh_hi);
                use_scaling = 1;
        }

NTP daemon in the guest can correct this difference (NTP can correct upto 500ppm).

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Message-Id: <20200616114741.GA298183@fuller.cnet>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/x86.c

index 29d9b078ce69304d36c9b09cd10349fa00e32b3f..3b92db412335d38ecc4367409e9c6c98bb9af55b 100644 (file)
@@ -4603,7 +4603,8 @@ long kvm_arch_vcpu_ioctl(struct file *filp,
                r = -EINVAL;
                user_tsc_khz = (u32)arg;
 
-               if (user_tsc_khz >= kvm_max_guest_tsc_khz)
+               if (kvm_has_tsc_control &&
+                   user_tsc_khz >= kvm_max_guest_tsc_khz)
                        goto out;
 
                if (user_tsc_khz == 0)